A major problem that limits clinical islet transplantation is insufficient information on the critical quantity of islets needed to reverse insulin dependence. To address this problem, we have previously identified the critical number of purified islets of known size that consistently induced normoglycemia in a large mammal model of type I diabetes. In the present studies, we found that the dose-response relationship between the volume of islets transplanted and consistent normoglycemia in dogs corresponded to greater than 4.1 microliters islet tissue per kilogram body weight. The functional outcome of similar quantities of purified islets was then examined after autoimplantation into splenic or hepatic sites and after splenic implantation by venous reflux or pulp injection. During prolonged follow-up, four of five initially normoglycemic recipients of intrahepatic islets became hyperglycemic within 1 year and the other failed at 26 months. In contrast, function of intrasplenic islets was more durable with delayed onset of hyperglycemia observed in only two of six grafts at 13 and 18 months. Sustained normoglycemia was induced by splenic venous reflux of islets in six of seven dogs, but intrapulp injection succeeded in only two of six. Islet allografts implanted to the spleen (n = 10) or to the kidney capsule (n = 6) of Cyclosporine-treated recipients induced normoglycemia in all, but sustained function ensued in only the intrasplenic group when the islet mass was augmented by 40%. These data define the critical islet volume needed to induce normoglycemia in a large mammal. Islets implanted by venous reflux to spleen provide more durable long-term function than the liver.(ABSTRACT TRUNCATED AT 250 WORDS)